Search

The question of how to generate high-utility mutations, to be used for testing purposes, forms a key challenge in mutation testing literature. %Existing approaches rely either on human-specified syntactic rules or learning-based approaches, all of which produce large numbers of redundant mutants. Large Language Models (LLMs) have shown great potential in code-related tasks but their utility in mutation testing remains unexplored. To this end, we systematically investigate the performance of LLMs in generating effective mutations w.r.t. to their usability, fault detection potential, and relationship with real bugs. In particular, we perform a large-scale empirical study involving 4 LLMs, including both open- and closed-source models, and 440 real bugs on two Java benchmarks. We find that compared to existing approaches, LLMs generate more diverse mutations that are behaviorally closer to real bugs, which leads to approximately 18% higher fault detection than current approaches (i.e., 87% vs. 69%) in a newly collected set of bugs, purposely selected for evaluating learning-based approaches, i.e., mitigating potential data leakage concerns. Additionally, we explore alternative prompt engineering strategies and the root causes of uncompilable mutations, produced by the LLMs, and provide valuable insights for the use of LLMs in the context of mutation testing., Comment: 13 pages, 3 figures

Pre-trained code models lead the era of code intelligence. Many models have been designed with impressive performance recently. However, one important problem, data augmentation for code data that automatically helps developers prepare training data lacks study in the field of code learning. In this paper, we introduce a general data augmentation framework, GenCode, to enhance the training of code understanding models. GenCode follows a generation-and-selection paradigm to prepare useful training codes. Specifically, it uses code transformation techniques to generate new code candidates first and then selects important ones as the training data by importance metrics. To evaluate the effectiveness of GenCode with a general importance metric -- loss value, we conduct experiments on four code understanding tasks (e.g., code clone detection) and three pre-trained code models (e.g., CodeT5). Compared to the state-of-the-art (SOTA) code augmentation method, MixCode, GenCode produces code models with 2.92% higher accuracy and 4.90% robustness on average.

RESTful APIs are popular web services, requiring documentation to ease their comprehension, reusability and testing practices. The OpenAPI Specification (OAS) is a widely adopted and machine-readable format used to document such APIs. However, manually documenting RESTful APIs is a time-consuming and error-prone task, resulting in unavailable, incomplete, or imprecise documentation. As RESTful API testing tools require an OpenAPI specification as input, insufficient or informal documentation hampers testing quality. Recently, Large Language Models (LLMs) have demonstrated exceptional abilities to automate tasks based on their colossal training data. Accordingly, such capabilities could be utilized to assist the documentation and testing process of RESTful APIs. In this paper, we present RESTSpecIT, the first automated RESTful API specification inference and black-box testing approach leveraging LLMs. The approach requires minimal user input compared to state-of-the-art RESTful API inference and testing tools; Given an API name and an LLM key, HTTP requests are generated and mutated with data returned by the LLM. By sending the requests to the API endpoint, HTTP responses can be analyzed for inference and testing purposes. RESTSpecIT utilizes an in-context prompt masking strategy, requiring no model fine-tuning. Our evaluation demonstrates that RESTSpecIT is capable of: (1) inferring specifications with 85.05% of GET routes and 81.05% of query parameters found on average, (2) discovering undocumented and valid routes and parameters, and (3) uncovering server errors in RESTful APIs. Inferred specifications can also be used as testing tool inputs.

Much research on Machine Learning testing relies on empirical studies that evaluate and show their potential. However, in this context empirical results are sensitive to a number of parameters that can adversely impact the results of the experiments and potentially lead to wrong conclusions (Type I errors, i.e., incorrectly rejecting the Null Hypothesis). To this end, we survey the related literature and identify 10 commonly adopted empirical evaluation hazards that may significantly impact experimental results. We then perform a sensitivity analysis on 30 influential studies that were published in top-tier SE venues, against our hazard set and demonstrate their criticality. Our findings indicate that all 10 hazards we identify have the potential to invalidate experimental findings, such as those made by the related literature, and should be handled properly. Going a step further, we propose a point set of 10 good empirical practices that has the potential to mitigate the impact of the hazards. We believe our work forms the first step towards raising awareness of the common pitfalls and good practices within the software engineering community and hopefully contribute towards setting particular expectations for empirical research in the field of deep learning testing.

Large Language Models (LLMs) have shown promise in multiple software engineering tasks including code generation, program repair, code summarisation, and test generation. Fault localisation is instrumental in enabling automated debugging and repair of programs and was prominently featured as a highlight during the launch event of ChatGPT-4. Nevertheless, the performance of LLMs compared to state-of-the-art methods, as well as the impact of prompt design and context length on their efficacy, remains unclear. To fill this gap, this paper presents an in-depth investigation into the capability of ChatGPT-3.5 and ChatGPT-4, the two state-of-the-art LLMs, on fault localisation. Using the widely-adopted large-scale Defects4J dataset, we compare the two LLMs with the existing fault localisation techniques. We also investigate the consistency of LLMs in fault localisation, as well as how prompt engineering and the length of code context affect the fault localisation effectiveness. Our findings demonstrate that within function-level context, ChatGPT-4 outperforms all the existing fault localisation methods. Additional error logs can further improve ChatGPT models' localisation accuracy and consistency, with an average 46.9% higher accuracy over the state-of-the-art baseline SmartFL on the Defects4J dataset in terms of TOP-1 metric. However, when the code context of the Defects4J dataset expands to the class-level, ChatGPT-4's performance suffers a significant drop, with 49.9% lower accuracy than SmartFL under TOP-1 metric. These observations indicate that although ChatGPT can effectively localise faults under specific conditions, limitations are evident. Further research is needed to fully harness the potential of LLMs like ChatGPT for practical fault localisation applications.

Testing deep learning-based systems is crucial but challenging due to the required time and labor for labeling collected raw data. To alleviate the labeling effort, multiple test selection methods have been proposed where only a subset of test data needs to be labeled while satisfying testing requirements. However, we observe that such methods with reported promising results are only evaluated under simple scenarios, e.g., testing on original test data. This brings a question to us: are they always reliable? In this paper, we explore when and to what extent test selection methods fail for testing. Specifically, first, we identify potential pitfalls of 11 selection methods from top-tier venues based on their construction. Second, we conduct a study on five datasets with two model architectures per dataset to empirically confirm the existence of these pitfalls. Furthermore, we demonstrate how pitfalls can break the reliability of these methods. Concretely, methods for fault detection suffer from test data that are: 1) correctly classified but uncertain, or 2) misclassified but confident. Remarkably, the test relative coverage achieved by such methods drops by up to 86.85%. On the other hand, methods for performance estimation are sensitive to the choice of intermediate-layer output. The effectiveness of such methods can be even worse than random selection when using an inappropriate layer., Comment: 12 pages

The costly human effort required to prepare the training data of machine learning (ML) models hinders their practical development and usage in software engineering (ML4Code), especially for those with limited budgets. Therefore, efficiently training models of code with less human effort has become an emergent problem. Active learning is such a technique to address this issue that allows developers to train a model with reduced data while producing models with desired performance, which has been well studied in computer vision and natural language processing domains. Unfortunately, there is no such work that explores the effectiveness of active learning for code models. In this paper, we bridge this gap by building the first benchmark to study this critical problem - active code learning. Specifically, we collect 11 acquisition functions~(which are used for data selection in active learning) from existing works and adapt them for code-related tasks. Then, we conduct an empirical study to check whether these acquisition functions maintain performance for code data. The results demonstrate that feature selection highly affects active learning and using output vectors to select data is the best choice. For the code summarization task, active code learning is ineffective which produces models with over a 29.64\% gap compared to the expected performance. Furthermore, we explore future directions of active code learning with an exploratory study. We propose to replace distance calculation methods with evaluation metrics and find a correlation between these evaluation-based distance methods and the performance of code models., Comment: 12 pages, ongoing work

Asynchronous waits are one of the most prevalent root causes of flaky tests and a major time-influential factor of web application testing. To investigate the characteristics of asynchronous wait flaky tests and their fixes in web testing, we build a dataset of 49 reproducible flaky tests, from 26 open-source projects, caused by asynchronous waits, along with their corresponding developer-written fixes. Our study of these flaky tests reveals that in approximately 63% of them (31 out of 49), developers addressed Asynchronous Wait flaky tests by adapting the wait time, even for cases where the root causes lie elsewhere. Based on this finding, we propose TRaf, an automated time-based repair method for asynchronous wait flaky tests in web applications. TRaf tackles the flakiness issues by suggesting a proper waiting time for each asynchronous call in a web application, using code similarity and past change history. The core insight is that as developers often make similar mistakes more than once, hints for the efficient wait time exist in the current or past codebase. Our analysis shows that TRaf can suggest a shorter wait time to resolve the test flakiness compared to developer-written fixes, reducing the test execution time by 11.1%. With additional dynamic tuning of the new wait time, TRaf further reduces the execution time by 20.2%.

The next era of program understanding is being propelled by the use of machine learning to solve software problems. Recent studies have shown surprising results of source code learning, which applies deep neural networks (DNNs) to various critical software tasks, e.g., bug detection and clone detection. This success can be greatly attributed to the utilization of massive high-quality training data, and in practice, data augmentation, which is a technique used to produce additional training data, has been widely adopted in various domains, such as computer vision. However, in source code learning, data augmentation has not been extensively studied, and existing practice is limited to simple syntax-preserved methods, such as code refactoring. Essentially, source code is often represented in two ways, namely, sequentially as text data and structurally as graph data, when it is used as training data in source code learning. Inspired by these analogy relations, we take an early step to investigate whether data augmentation methods that are originally used for text and graphs are effective in improving the training quality of source code learning. To that end, we first collect and categorize data augmentation methods in the literature. Second, we conduct a comprehensive empirical study on four critical tasks and 11 DNN architectures to explore the effectiveness of 12 data augmentation methods (including code refactoring and 11 other methods for text and graph data). Our results identify the data augmentation methods that can produce more accurate and robust models for source code learning, including those based on mixup (e.g., SenMixup for texts and Manifold-Mixup for graphs), and those that slightly break the syntax of source code (e.g., random swap and random deletion for texts).

System goals are the statements that, in the context of software requirements specification, capture how the software should behave. Many times, the understanding of stakeholders on what the system should do, as captured in the goals, can lead to different problems, from clearly contradicting goals, to more subtle situations in which the satisfaction of some goals inhibits the satisfaction of others. These latter issues, called goal divergences, are the subject of goal conflict analysis, which consists of identifying, assessing, and resolving divergences, as part of a more general activity known as goal refinement. While there exist techniques that, when requirements are expressed formally, can automatically identify and assess goal conflicts, there is currently no automated approach to support engineers in resolving identified divergences. In this paper, we present ACoRe, the first approach that automatically proposes potential resolutions to goal conflicts, in requirements specifications formally captured using linear-time temporal logic. ACoRe systematically explores syntactic modifications of the conflicting specifications, aiming at obtaining resolutions that disable previously identified conflicts, while preserving specification consistency. ACoRe integrates modern multi-objective search algorithms (in particular, NSGA-III, WBGA, and AMOSA) to produce resolutions that maintain coherence with the original conflicting specification, by searching for specifications that are either syntactically or semantically similar to the original specification. We assess ACoRe on 25 requirements specifications taken from the literature. We show that ACoRe can successfully produce various conflict resolutions for each of the analyzed case studies, including resolutions that resemble specification repairs manually provided as part of conflict analyses.

With the increasing release of powerful language models trained on large code corpus (e.g. CodeBERT was trained on 6.4 million programs), a new family of mutation testing tools has arisen with the promise to generate more "natural" mutants in the sense that the mutated code aims at following the implicit rules and coding conventions typically produced by programmers. In this paper, we study to what extent the mutants produced by language models can semantically mimic the observable behavior of security-related vulnerabilities (a.k.a. Vulnerability-mimicking Mutants), so that designing test cases that are failed by these mutants will help in tackling mimicked vulnerabilities. Since analyzing and running mutants is computationally expensive, it is important to prioritize those mutants that are more likely to be vulnerability mimicking prior to any analysis or test execution. Taking this into account, we introduce VMMS, a machine learning based approach that automatically extracts the features from mutants and predicts the ones that mimic vulnerabilities. We conducted our experiments on a dataset of 45 vulnerabilities and found that 16.6% of the mutants fail one or more tests that are failed by 88.9% of the respective vulnerabilities. More precisely, 3.9% of the mutants from the entire mutant set are vulnerability-mimicking mutants that mimic 55.6% of the vulnerabilities. Despite the scarcity, VMMS predicts vulnerability-mimicking mutants with 0.63 MCC, 0.80 Precision, and 0.51 Recall, demonstrating that the features of vulnerability-mimicking mutants can be automatically learned by machine learning models to statically predict these without the need of investing effort in defining such features., Comment: arXiv admin note: text overlap with arXiv:2301.12284

Flaky tests are tests that pass and fail on different executions of the same version of a program under test. They waste valuable developer time by making developers investigate false alerts (flaky test failures). To deal with this problem, many prediction methods that identify flaky tests have been proposed. While promising, the actual utility of these methods remains unclear since they have not been evaluated within a continuous integration (CI) process. In particular, it remains unclear what is the impact of missed faults, i.e., the consideration of fault-triggering test failures as flaky, at different CI cycles. To fill this gap, we apply state-of-the-art flakiness prediction methods at the Chromium CI and check their performance. Perhaps surprisingly, we find that, despite the high precision (99.2%) of the methods, their application leads to numerous faults missed, approximately 76.2% of all regression faults. To explain this result, we analyse the fault-triggering failures and show that flaky tests have a strong fault-revealing capability, i.e., they reveal more than 1/3 of all regression faults, indicating an inherent limitation of all methods focusing on identifying flaky tests, instead of flaky test failures. Going a step further, we build failure-focused prediction methods and optimize them by considering new features. Interestingly, we find that these methods perform better than the test-focused ones, with an MCC increasing from 0.20 to 0.42. Overall, our findings imply that on the one hand future research should focus on predicting flaky test failures instead of flaky tests and the need for adopting more thorough experimental methodologies when evaluating flakiness prediction methods, on the other.

While leveraging additional training data is well established to improve adversarial robustness, it incurs the unavoidable cost of data collection and the heavy computation to train models. To mitigate the costs, we propose Guided Adversarial Training (GAT), a novel adversarial training technique that exploits auxiliary tasks under a limited set of training data. Our approach extends single-task models into multi-task models during the min-max optimization of adversarial training, and drives the loss optimization with a regularization of the gradient curvature across multiple tasks. GAT leverages two types of auxiliary tasks: self-supervised tasks, where the labels are generated automatically, and domain-knowledge tasks, where human experts provide additional labels. Experimentally, GAT increases the robust AUC of CheXpert medical imaging dataset from 50% to 83% and On CIFAR-10, GAT outperforms eight state-of-the-art adversarial training and achieves 56.21% robust accuracy with Resnet-50. Overall, we demonstrate that guided multi-task learning is an actionable and promising avenue to push further the boundaries of model robustness.

In this paper, we propose an assertion-based approach to capture software evolution, through the notion of commit-relevant specification. A commit-relevant specification summarises the program properties that have changed as a consequence of a commit (understood as a specific software modification), via two sets of assertions, the delta-added assertions, properties that did not hold in the pre-commit version but hold on the post-commit, and the delta-removed assertions, those that were valid in the pre-commit, but no longer hold after the code change. We also present DeltaSpec, an approach that combines test generation and dynamic specification inference to automatically compute commit-relevant specifications from given commits. We evaluate DeltaSpec on two datasets that include a total of 57 commits (63 classes and 797 methods). We show that commit-relevant assertions can precisely describe the semantic deltas of code changes, providing a useful mechanism for validating the behavioural evolution of software. We also show that DeltaSpec can infer 88% of the manually written commit-relevant assertions expressible in the language supported by the tool. Moreover, our experiments demonstrate that DeltaSpec's inferred assertions are effective to detect regression faults. More precisely, we show that commit-relevant assertions can detect, on average, 78.3% of the artificially seeded faults that interact with the code changes. We also show that assertions in the delta are 58.3% more effective in detecting commit-relevant mutants than assertions outside the delta, and that it takes on average 169% fewer assertions when these are commit-relevant, compared to using general valid assertions, to achieve a same commit-relevant mutation score.

Specification inference techniques aim at (automatically) inferring a set of assertions that capture the exhibited software behaviour by generating and filtering assertions through dynamic test executions and mutation testing. Although powerful, such techniques are computationally expensive due to a large number of assertions, test cases and mutated versions that need to be executed. To overcome this issue, we demonstrate that a small subset, i.e., 12.95% of the mutants used by mutation testing tools is sufficient for assertion inference, this subset is significantly different, i.e., 71.59% different from the subsuming mutant set that is frequently cited by mutation testing literature, and can be statically approximated through a learning based method. In particular, we propose AIMS, an approach that selects Assertion Inferring Mutants, i.e., a set of mutants that are well-suited for assertion inference, with 0.58 MCC, 0.79 Precision, and 0.49 Recall. We evaluate AIMS on 46 programs and demonstrate that it has comparable inference capabilities with full mutation analysis (misses 12.49% of assertions) while significantly limiting execution cost (runs 46.29 times faster). A comparison with randomly selected sets of mutants, shows the superiority of AIMS by inferring 36% more assertions while requiring approximately equal amount of execution time. We also show that AIMS 's inferring capabilities are almost complete as it infers 96.15% of ground truth assertions, (i.e., a complete set of assertions that were manually constructed) while Random Mutant Selection infers 19.23% of them. More importantly, AIMS enables assertion inference techniques to scale on subjects where full mutation testing is prohibitively expensive and Random Mutant Selection does not lead to any assertion.

Mutation testing is an established fault-based testing technique. It operates by seeding faults into the programs under test and asking developers to write tests that reveal these faults. These tests have the potential to reveal a large number of faults -- those that couple with the seeded ones -- and thus are deemed important. To this end, mutation testing should seed faults that are both "natural" in a sense easily understood by developers and strong (have high chances to reveal faults). To achieve this we propose using pre-trained generative language models (i.e. CodeBERT) that have the ability to produce developer-like code that operates similarly, but not exactly, as the target code. This means that the models have the ability to seed natural faults, thereby offering opportunities to perform mutation testing. We realise this idea by implementing $\mu$BERT, a mutation testing technique that performs mutation testing using CodeBert and empirically evaluated it using 689 faulty program versions. Our results show that the fault revelation ability of $\mu$BERT is higher than that of a state-of-the-art mutation testing (PiTest), yielding tests that have up to 17% higher fault detection potential than that of PiTest. Moreover, we observe that $\mu$BERT can complement PiTest, being able to detect 47 bugs missed by PiTest, while at the same time, PiTest can find 13 bugs missed by $\mu$BERT.

Mutation testing has been demonstrated to be one of the most powerful fault-revealing tools in the tester's tool kit. Much previous work implicitly assumed it to be sufficient to re-compute mutant suites per release. Sadly, this makes mutation results inconsistent; mutant scores from each release cannot be directly compared, making it harder to measure test improvement. Furthermore, regular code change means that a mutant suite's relevance will naturally degrade over time. We measure this degradation in relevance for 143,500 mutants in 4 non-trivial systems finding that, on overage, 52% degrade. We introduce a mutant brittleness measure and use it to audit software systems and their mutation suites. We also demonstrate how consistent-by-construction long-standing mutant suites can be identified with a 10x improvement in mutant relevance over an arbitrary test suite. Our results indicate that the research community should avoid the re-computation of mutant suites and focus, instead, on long-standing mutants, thereby improving the consistency and relevance of mutation testing.

Graph neural network (GNN)-based graph learning has been popular in natural language and programming language processing, particularly in text and source code classification. Typically, GNNs are constructed by incorporating alternating layers which learn transformations of graph node features, along with graph pooling layers that use graph pooling operators (e.g., Max-pooling) to effectively reduce the number of nodes while preserving the semantic information of the graph. Recently, to enhance GNNs in graph learning tasks, Manifold-Mixup, a data augmentation technique that produces synthetic graph data by linearly mixing a pair of graph data and their labels, has been widely adopted. However, the performance of Manifold-Mixup can be highly affected by graph pooling operators, and there have not been many studies that are dedicated to uncovering such affection. To bridge this gap, we take an early step to explore how graph pooling operators affect the performance of Mixup-based graph learning. To that end, we conduct a comprehensive empirical study by applying Manifold-Mixup to a formal characterization of graph pooling based on 11 graph pooling operations (9 hybrid pooling operators, 2 non-hybrid pooling operators). The experimental results on both natural language datasets (Gossipcop, Politifact) and programming language datasets (JAVA250, Python800) demonstrate that hybrid pooling operators are more effective for Manifold-Mixup than the standard Max-pooling and the state-of-the-art graph multiset transformer (GMT) pooling, in terms of producing more accurate and robust GNN models., Comment: Accepted by Journal of Systems and Software (JSS) 2024

Inspired by the great success of Deep Neural Networks (DNNs) in natural language processing (NLP), DNNs have been increasingly applied in source code analysis and attracted significant attention from the software engineering community. Due to its data-driven nature, a DNN model requires massive and high-quality labeled training data to achieve expert-level performance. Collecting such data is often not hard, but the labeling process is notoriously laborious. The task of DNN-based code analysis even worsens the situation because source code labeling also demands sophisticated expertise. Data augmentation has been a popular approach to supplement training data in domains such as computer vision and NLP. However, existing data augmentation approaches in code analysis adopt simple methods, such as data transformation and adversarial example generation, thus bringing limited performance superiority. In this paper, we propose a data augmentation approach MIXCODE that aims to effectively supplement valid training data, inspired by the recent advance named Mixup in computer vision. Specifically, we first utilize multiple code refactoring methods to generate transformed code that holds consistent labels with the original data. Then, we adapt the Mixup technique to mix the original code with the transformed code to augment the training data. We evaluate MIXCODE on two programming languages (Java and Python), two code tasks (problem classification and bug detection), four benchmark datasets (JAVA250, Python800, CodRep1, and Refactory), and seven model architectures (including two pre-trained models CodeBERT and GraphCodeBERT). Experimental results demonstrate that MIXCODE outperforms the baseline data augmentation approach by up to 6.24% in accuracy and 26.06% in robustness., Comment: Accepted by SANER 2023

Flaky tests are tests that yield different outcomes when run on the same version of a program. This non-deterministic behaviour plagues continuous integration with false signals, wasting developers' time and reducing their trust in test suites. Studies highlighted the importance of keeping tests flakiness-free. Recently, the research community has been pushing forward the detection of flaky tests by suggesting many static and dynamic approaches. While promising, those approaches mainly focus on classifying tests as flaky or not and, even when high performances are reported, it remains challenging to understand the cause of flakiness. This part is crucial for researchers and developers that aim to fix it. To help with the comprehension of a given flaky test, we propose FlakyCat, the first approach for classifying flaky tests based on their root cause category. FlakyCat relies on CodeBERT for code representation and leverages a Siamese network-based Few-Shot learning method to train a multi-class classifier with few data. We train and evaluate FlakyCat on a set of 343 flaky tests collected from open-source Java projects. Our evaluation shows that FlakyCat categorises flaky tests accurately, with a weighted F1 score of 70%. Furthermore, we investigate the performance of our approach for each category, revealing that Async waits, Unordered collections and Time-related flaky tests are accurately classified, while Concurrency-related flaky tests are more challenging to predict. Finally, to facilitate the comprehension of FlakyCat's predictions, we present a new technique for CodeBERT-based model interpretability that highlights code statements influencing the categorization.

Much of software-engineering research relies on the naturalness of code, the fact that code, in small code snippets, is repetitive and can be predicted using statistical language models like n-gram. Although powerful, training such models on large code corpus is tedious, time-consuming and sensitive to code patterns (and practices) encountered during training. Consequently, these models are often trained on a small corpora and estimate the language naturalness that is relative to a specific style of programming or type of project. To overcome these issues, we propose using pre-trained language models to infer code naturalness. Pre-trained models are often built on big data, are easy to use in an out-of-the-box way and include powerful learning associations mechanisms. Our key idea is to quantify code naturalness through its predictability, by using state-of-the-art generative pre-trained language models. To this end, we infer naturalness by masking (omitting) code tokens, one at a time, of code-sequences, and checking the models' ability to predict them. To this end, we evaluate three different predictability metrics; a) measuring the number of exact matches of the predictions, b) computing the embedding similarity between the original and predicted code, i.e., similarity at the vector space, and c) computing the confidence of the model when doing the token completion task irrespective of the outcome. We implement this workflow, named CodeBERT-nt, and evaluate its capability to prioritize buggy lines over non-buggy ones when ranking code based on its naturalness. Our results, on 2510 buggy versions of 40 projects from the SmartShark dataset, show that CodeBERT-nt outperforms both, random-uniform and complexity-based ranking techniques, and yields comparable results (slightly better) than the n-gram models.

We propose transferability from Large Geometric Vicinity (LGV), a new technique to increase the transferability of black-box adversarial attacks. LGV starts from a pretrained surrogate model and collects multiple weight sets from a few additional training epochs with a constant and high learning rate. LGV exploits two geometric properties that we relate to transferability. First, models that belong to a wider weight optimum are better surrogates. Second, we identify a subspace able to generate an effective surrogate ensemble among this wider optimum. Through extensive experiments, we show that LGV alone outperforms all (combinations of) four established test-time transformations by 1.8 to 59.9 percentage points. Our findings shed new light on the importance of the geometry of the weight space to explain the transferability of adversarial examples., Comment: Accepted at ECCV 2022

Vulnerability prediction refers to the problem of identifying system components that are most likely to be vulnerable. Typically, this problem is tackled by training binary classifiers on historical data. Unfortunately, recent research has shown that such approaches underperform due to the following two reasons: a) the imbalanced nature of the problem, and b) the inherently noisy historical data, i.e., most vulnerabilities are discovered much later than they are introduced. This misleads classifiers as they learn to recognize actual vulnerable components as non-vulnerable. To tackle these issues, we propose TROVON, a technique that learns from known vulnerable components rather than from vulnerable and non-vulnerable components, as typically performed. We perform this by contrasting the known vulnerable, and their respective fixed components. This way, TROVON manages to learn from the things we know, i.e., vulnerabilities, hence reducing the effects of noisy and unbalanced data. We evaluate TROVON by comparing it with existing techniques on three security-critical open source systems, i.e., Linux Kernel, OpenSSL, and Wireshark, with historical vulnerabilities that have been reported in the National Vulnerability Database (NVD). Our evaluation demonstrates that the prediction capability of TROVON significantly outperforms existing vulnerability prediction techniques such as Software Metrics, Imports, Function Calls, Text Mining, Devign, LSTM, and LSTM-RF with an improvement of 40.84% in Matthews Correlation Coefficient (MCC) score under Clean Training Data Settings, and an improvement of 35.52% under Realistic Training Data Settings., Comment: Please do not consider this new version of the article for citations. The article (with its previous versions) is already available here: arXiv:2012.11701

Deep learning (DL) plays a more and more important role in our daily life due to its competitive performance in industrial application domains. As the core of DL-enabled systems, deep neural networks (DNNs) need to be carefully evaluated to ensure the produced models match the expected requirements. In practice, the \emph{de facto standard} to assess the quality of DNNs in the industry is to check their performance (accuracy) on a collected set of labeled test data. However, preparing such labeled data is often not easy partly because of the huge labeling effort, i.e., data labeling is labor-intensive, especially with the massive new incoming unlabeled data every day. Recent studies show that test selection for DNN is a promising direction that tackles this issue by selecting minimal representative data to label and using these data to assess the model. However, it still requires human effort and cannot be automatic. In this paper, we propose a novel technique, named \textit{Aries}, that can estimate the performance of DNNs on new unlabeled data using only the information obtained from the original test data. The key insight behind our technique is that the model should have similar prediction accuracy on the data which have similar distances to the decision boundary. We performed a large-scale evaluation of our technique on two famous datasets, CIFAR-10 and Tiny-ImageNet, four widely studied DNN models including ResNet101 and DenseNet121, and 13 types of data transformation methods. Results show that the estimated accuracy by \textit{Aries} is only 0.03\% -- 2.60\% off the true accuracy. Besides, \textit{Aries} also outperforms the state-of-the-art labeling-free methods in 50 out of 52 cases and selection-labeling-based methods in 96 out of 128 cases., Comment: accepted to ICSE'23, preprint version

Flaky tests are defined as tests that manifest non-deterministic behaviour by passing and failing intermittently for the same version of the code. These tests cripple continuous integration with false alerts that waste developers' time and break their trust in regression testing. To mitigate the effects of flakiness, both researchers and industrial experts proposed strategies and tools to detect and isolate flaky tests. However, flaky tests are rarely fixed as developers struggle to localise and understand their causes. Additionally, developers working with large codebases often need to know the sources of non-determinism to preserve code quality, i.e., avoid introducing technical debt linked with non-deterministic behaviour, and to avoid introducing new flaky tests. To aid with these tasks, we propose re-targeting Fault Localisation techniques to the flaky component localisation problem, i.e., pinpointing program classes that cause the non-deterministic behaviour of flaky tests. In particular, we employ Spectrum-Based Fault Localisation (SBFL), a coverage-based fault localisation technique commonly adopted for its simplicity and effectiveness. We also utilise other data sources, such as change history and static code metrics, to further improve the localisation. Our results show that augmenting SBFL with change and code metrics ranks flaky classes in the top-1 and top-5 suggestions, in 26% and 47% of the cases. Overall, we successfully reduced the average number of classes inspected to locate the first flaky class to 19% of the total number of classes covered by flaky tests. Our results also show that localisation methods are effective in major flakiness categories, such as concurrency and asynchronous waits, indicating their general ability to identify flaky components., Comment: Accepted at the 38th IEEE International Conference on Software Maintenance and Evolution (ICSME)

Distribution shift has been a longstanding challenge for the reliable deployment of deep learning (DL) models due to unexpected accuracy degradation. Although DL has been becoming a driving force for large-scale source code analysis in the big code era, limited progress has been made on distribution shift analysis and benchmarking for source code tasks. To fill this gap, this paper initiates to propose CodeS, a distribution shift benchmark dataset, for source code learning. Specifically, CodeS supports two programming languages (Java and Python) and five shift types (task, programmer, time-stamp, token, and concrete syntax tree). Extensive experiments based on CodeS reveal that 1) out-of-distribution detectors from other domains (e.g., computer vision) do not generalize to source code, 2) all code classification models suffer from distribution shifts, 3) representation-based shifts have a higher impact on the model than others, and 4) pre-trained bimodal models are relatively more resistant to distribution shifts., Comment: accepted by ICSE'23-NIER

In the last decade, researchers have studied fairness as a software property. In particular, how to engineer fair software systems? This includes specifying, designing, and validating fairness properties. However, the landscape of works addressing bias as a software engineering concern is unclear, i.e., techniques and studies that analyze the fairness properties of learning-based software. In this work, we provide a clear view of the state-of-the-art in software fairness analysis. To this end, we collect, categorize and conduct an in-depth analysis of 164 publications investigating the fairness of learning-based software systems. Specifically, we study the evaluated fairness measure, the studied tasks, the type of fairness analysis, the main idea of the proposed approaches, and the access level (e.g., black, white, or grey box). Our findings include the following: (1) Fairness concerns (such as fairness specification and requirements engineering) are under-studied; (2) Fairness measures such as conditional, sequential, and intersectional fairness are under-explored; (3) Unstructured datasets (e.g., audio, image, and text) are barely studied for fairness analysis; and (4) Software fairness analysis techniques hardly employ white-box, in-processing machine learning (ML) analysis methods. In summary, we observed several open challenges including the need to study intersectional/sequential bias, policy-based bias handling, and human-in-the-loop, socio-technical bias mitigation.

Deep Neural Networks (DNNs) have gained considerable attention in the past decades due to their astounding performance in different applications, such as natural language modeling, self-driving assistance, and source code understanding. With rapid exploration, more and more complex DNN architectures have been proposed along with huge pre-trained model parameters. The common way to use such DNN models in user-friendly devices (e.g., mobile phones) is to perform model compression before deployment. However, recent research has demonstrated that model compression, e.g., model quantization, yields accuracy degradation as well as outputs disagreements when tested on unseen data. Since the unseen data always include distribution shifts and often appear in the wild, the quality and reliability of quantized models are not ensured. In this paper, we conduct a comprehensive study to characterize and help users understand the behaviors of quantized models. Our study considers 4 datasets spanning from image to text, 8 DNN architectures including feed-forward neural networks and recurrent neural networks, and 42 shifted sets with both synthetic and natural distribution shifts. The results reveal that 1) data with distribution shifts happen more disagreements than without. 2) Quantization-aware training can produce more stable models than standard, adversarial, and Mixup training. 3) Disagreements often have closer top-1 and top-2 output probabilities, and $Margin$ is a better indicator than the other uncertainty metrics to distinguish disagreements. 4) Retraining with disagreements has limited efficiency in removing disagreements. We opensource our code and models as a new benchmark for further studying the quantized models., Comment: 12 pages

Applying deep learning to science is a new trend in recent years which leads DL engineering to become an important problem. Although training data preparation, model architecture design, and model training are the normal processes to build DL models, all of them are complex and costly. Therefore, reusing the open-sourced pre-trained model is a practical way to bypass this hurdle for developers. Given a specific task, developers can collect massive pre-trained deep neural networks from public sources for re-using. However, testing the performance (e.g., accuracy and robustness) of multiple DNNs and recommending which model should be used is challenging regarding the scarcity of labeled data and the demand for domain expertise. In this paper, we propose a labeling-free (LaF) model selection approach to overcome the limitations of labeling efforts for automated model reusing. The main idea is to statistically learn a Bayesian model to infer the models' specialty only based on predicted labels. We evaluate LaF using 9 benchmark datasets including image, text, and source code, and 165 DNNs, considering both the accuracy and robustness of models. The experimental results demonstrate that LaF outperforms the baseline methods by up to 0.74 and 0.53 on Spearman's correlation and Kendall's $\tau$, respectively., Comment: 22 pages

Many software engineering techniques, such as fault localization, operate based on relevance relationships between tests and code. These relationships are often inferred through the use of dynamic test execution information (test execution traces) that approximate the link between relevant code units and asserted, by the tests, program behaviour. Unfortunately, in practice dynamic information is not always available due to the overheads introduced by the instrumentation or the nature of the production environments. To deal with this issue, we propose CEMENT, a static technique that automatically infers such test and code relationships given the projects' evolution. The key idea is that developers make relevant changes on test and code units at the same period of time, i.e., co-evolution of tests and code units reflects a probable link between them. We evaluate CEMENT on 15 open source projects and show that it indeed captures relevant links. Additionally, we perform a fault localization case study where we compare CEMENT with an existing Information Retrieval-based Fault Localization (IRFL) technique and show that it achieves comparable performance. A further analysis of our results reveals a small overlap between the faults successfully localized by the two approaches suggesting complementarity. In particular, out of the 39 successfully localized faults, two are common while CEMENT and IRFL localize 16 and 21. These results demonstrate that test and code evolutionary coupling can effectively support test and debugging activities.

We introduce $\mu$BERT, a mutation testing tool that uses a pre-trained language model (CodeBERT) to generate mutants. This is done by masking a token from the expression given as input and using CodeBERT to predict it. Thus, the mutants are generated by replacing the masked tokens with the predicted ones. We evaluate $\mu$BERT on 40 real faults from Defects4J and show that it can detect 27 out of the 40 faults, while the baseline (PiTest) detects 26 of them. We also show that $\mu$BERT can be 2 times more cost-effective than PiTest, when the same number of mutants are analysed. Additionally, we evaluate the impact of $\mu$BERT's mutants when used by program assertion inference techniques, and show that they can help in producing better specifications. Finally, we discuss about the quality and naturalness of some interesting mutants produced by $\mu$BERT during our experimental evaluation.

When software evolves, opportunities for introducing faults appear. Therefore, it is important to test the evolved program behaviors during each evolution cycle. We conduct an exploratory study to investigate the properties of commit-relevant mutants, i.e., the test elements of commit-aware mutation testing, by offering a general definition and an experimental approach to identify them. We thus, aim at investigating the prevalence, location, comparative advantages of commit-aware mutation testing over time (i.e., the program evolution) and the predictive power of several commit-related features to understand the essential properties for its best-effort application case. Our approach utilizes the impact of mutants and the effects of one mutant on another in capturing and analyzing the implicit interactions between the changed and unchanged code parts. The study analyses millions of mutants (over 10 million), 288 commits, five (5) different open-source software projects involving over 68,213 CPU days of computation and sets a ground truth where we perform our analysis. Our analysis shows that commit-relevant mutants are located mainly outside of program commit change (81%), while an effective selection of commit-relevant mutants can reduce the number of mutants by up to 93%. In addition, we demonstrate that commit relevant mutation testing is significantly more effective and efficient than state-of-the-art baselines. Our analysis of the predictive power of mutants and commit-related features in predicting commit-relevant mutants found that most proxy features do not reliably predict commit-relevant mutants. This empirical study highlights the properties of commit-relevant mutants and demonstrates the importance of identifying and selecting commit-relevant mutants when testing evolving software systems.

Fault seeding is typically used in controlled studies to evaluate and compare test techniques. Central to these techniques lies the hypothesis that artificially seeded faults involve some form of realistic properties and thus provide realistic experimental results. In an attempt to strengthen realism, a recent line of research uses advanced machine learning techniques, such as deep learning and Natural Language Processing (NLP), to seed faults that look like (syntactically) real ones, implying that fault realism is related to syntactic similarity. This raises the question of whether seeding syntactically similar faults indeed results in semantically similar faults and more generally whether syntactically dissimilar faults are far away (semantically) from the real ones. We answer this question by employing 4 fault-seeding techniques (PiTest - a popular mutation testing tool, IBIR - a tool with manually crafted fault patterns, DeepMutation - a learning-based fault seeded framework and CodeBERT - a novel mutation testing tool that use code embeddings) and demonstrate that syntactic similarity does not reflect semantic similarity. We also show that 60%, 47%, 43%, and 7% of the real faults of Defects4J V2 are semantically resembled by CodeBERT, PiTest, IBIR, and DeepMutation faults. We then perform an objective comparison between the techniques and find that CodeBERT and PiTest have similar fault detection capabilities that subsume IBIR and DeepMutation, and that IBIR is the most cost-effective technique. Moreover, the overall fault detection of PiTest, CodeBERT, IBIR, and DeepMutation was, on average, 54%, 53%, 37%, and 7%.

Mutation testing research has indicated that a major part of its application cost is due to the large number of low utility mutants that it introduces. Although previous research has identified this issue, no previous study has proposed any effective solution to the problem. Thus, it remains unclear how to mutate and test a given piece of code in a best effort way, i.e., achieving a good trade-off between invested effort and test effectiveness. To achieve this, we propose Cerebro, a machine learning approach that statically selects subsuming mutants, i.e., the set of mutants that resides on the top of the subsumption hierarchy, based on the mutants' surrounding code context. We evaluate Cerebro using 48 and 10 programs written in C and Java, respectively, and demonstrate that it preserves the mutation testing benefits while limiting application cost, i.e., reduces all cost application factors such as equivalent mutants, mutant executions, and the mutants requiring analysis. We demonstrate that Cerebro has strong inter-project prediction ability, which is significantly higher than two baseline methods, i.e., supervised learning on features proposed by state-of-the-art, and random mutant selection. More importantly, our results show that Cerebro's selected mutants lead to strong tests that are respectively capable of killing 2 times higher than the number of subsuming mutants killed by the baselines when selecting the same number of mutants. At the same time, Cerebro reduces the cost-related factors, as it selects, on average, 68% fewer equivalent mutants, while requiring 90% fewer test executions than the baselines.

Test flakiness forms a major testing concern. Flaky tests manifest non-deterministic outcomes that cripple continuous integration and lead developers to investigate false alerts. Industrial reports indicate that on a large scale, the accrual of flaky tests breaks the trust in test suites and entails significant computational cost. To alleviate this, practitioners are constrained to identify flaky tests and investigate their impact. To shed light on such mitigation mechanisms, we interview 14 practitioners with the aim to identify (i) the sources of flakiness within the testing ecosystem, (ii) the impacts of flakiness, (iii) the measures adopted by practitioners when addressing flakiness, and (iv) the automation opportunities for these measures. Our analysis shows that, besides the tests and code, flakiness stems from interactions between the system components, the testing infrastructure, and external factors. We also highlight the impact of flakiness on testing practices and product quality and show that the adoption of guidelines together with a stable infrastructure are key measures in mitigating the problem.

Active learning is an established technique to reduce the labeling cost to build high-quality machine learning models. A core component of active learning is the acquisition function that determines which data should be selected to annotate. State-of-the-art acquisition functions -- and more largely, active learning techniques -- have been designed to maximize the clean performance (e.g. accuracy) and have disregarded robustness, an important quality property that has received increasing attention. Active learning, therefore, produces models that are accurate but not robust. In this paper, we propose \emph{robust active learning}, an active learning process that integrates adversarial training -- the most established method to produce robust models. Via an empirical study on 11 acquisition functions, 4 datasets, 6 DNN architectures, and 15105 trained DNNs, we show that robust active learning can produce models with the robustness (accuracy on adversarial examples) ranging from 2.35\% to 63.85\%, whereas standard active learning systematically achieves negligible robustness (less than 0.20\%). Our study also reveals, however, that the acquisition functions that perform well on accuracy are worse than random sampling when it comes to robustness. We, therefore, examine the reasons behind this and devise a new acquisition function that targets both clean performance and robustness. Our acquisition function -- named density-based robust sampling with entropy (DRE) -- outperforms the other acquisition functions (including random) in terms of robustness by up to 24.40\% (3.84\% than random particularly), while remaining competitive on accuracy. Additionally, we prove that DRE is applicable as a test selection metric for model retraining and stands out from all compared functions by up to 8.21\% robustness., Comment: 10 pages

Code embedding is a keystone in the application of machine learning on several Software Engineering (SE) tasks. To effectively support a plethora of SE tasks, the embedding needs to capture program syntax and semantics in a way that is generic. To this end, we propose the first self-supervised pre-training approach (called GraphCode2Vec) which produces task-agnostic embedding of lexical and program dependence features. GraphCode2Vec achieves this via a synergistic combination of code analysis and Graph Neural Networks. GraphCode2Vec is generic, it allows pre-training, and it is applicable to several SE downstream tasks. We evaluate the effectiveness of GraphCode2Vec on four (4) tasks (method name prediction, solution classification, mutation testing and overfitted patch classification), and compare it with four (4) similarly generic code embedding baselines (Code2Seq, Code2Vec, CodeBERT, GraphCodeBERT) and 7 task-specific, learning-based methods. In particular, GraphCode2Vec is more effective than both generic and task-specific learning-based baselines. It is also complementary and comparable to GraphCodeBERT (a larger and more complex model). We also demonstrate through a probing and ablation study that GraphCode2Vec learns lexical and program dependence features and that self-supervised pre-training improves effectiveness.

Flakiness is a major concern in Software testing. Flaky tests pass and fail for the same version of a program and mislead developers who spend time and resources investigating test failures only to discover that they are false alerts. In practice, the defacto approach to address this concern is to rerun failing tests hoping that they would pass and manifest as false alerts. Nonetheless, completely filtering out false alerts may require a disproportionate number of reruns, and thus incurs important costs both computation and time-wise. As an alternative to reruns, we propose Fair, a novel, lightweight approach that classifies test failures into false alerts and legitimate failures. Fair relies on a classifier and a set of features from the failures and test artefacts. To build and evaluate our machine learning classifier, we use the continuous integration of the Chromium project. In particular, we collect the properties and artefacts of more than 1 million test failures from 2,000 builds. Our results show that Fair can accurately distinguish legitimate failures from false alerts, with an MCC up to 95%. Moreover, by studying different test categories: GUI, integration and unit tests, we show that Fair classifies failures accurately even when the number of failures is limited. Finally, we compare the costs of our approach to reruns and show that Fair could save up to 20 minutes of computation time per build.

Test smells are known as bad development practices that reflect poor design and implementation choices in software tests. Over the last decade, test smells were heavily studied to measure their prevalence and impacts on test maintainability. However, these studies focused mainly on the unit level and to this day, the work on system tests that interact with the System Under Test through a Graphical User Interface remains limited. To fill the gap, we conduct an exploratory analysis of test smells occurring in System User Interactive Tests (SUIT). First, based on a multi-vocal literature review, we propose a catalog of 35 SUIT-specific smells. Then, we conduct an empirical analysis to assess the prevalence and refactoring of these smells in 48 industrial test suites and 12 open-source projects. We show that the same type of smells tends to appear in industrial and open-source projects, but the symptoms are not addressed in the same way. Smells such as Obscure Test, Sneaky Checking, and Over Checking show symptoms in more than 70% of the tests. Yet refactoring actions are much less frequent with less than 50% of the affected tests ever undergoing refactoring. Interestingly, while refactoring actions are rare, some smells, such as Narcissistic, disappear through the removal of old symptomatic tests and the introduction of new tests not presenting such symptoms.

Vulnerability to adversarial attacks is a well-known weakness of Deep Neural networks. While most of the studies focus on single-task neural networks with computer vision datasets, very little research has considered complex multi-task models that are common in real applications. In this paper, we evaluate the design choices that impact the robustness of multi-task deep learning networks. We provide evidence that blindly adding auxiliary tasks, or weighing the tasks provides a false sense of robustness. Thereby, we tone down the claim made by previous research and study the different factors which may affect robustness. In particular, we show that the choice of the task to incorporate in the loss function are important factors that can be leveraged to yield more robust models.

The reactive synthesis problem consists of automatically producing correct-by-construction operational models of systems from high-level formal specifications of their behaviours. However, specifications are often unrealisable, meaning that no system can be synthesised from the specification. To deal with this problem, we present AuRUS, a search-based approach to repair unrealisable Linear-Time Temporal Logic (LTL) specifications. AuRUS aims at generating solutions that are similar to the original specifications by using the notions of syntactic and semantic similarities. Intuitively, the syntactic similarity measures the text similarity between the specifications, while the semantic similarity measures the number of behaviours preserved/removed by the candidate repair. We propose a new heuristic based on model counting to approximate semantic similarity. We empirically assess AuRUS on many unrealisable specifications taken from different benchmarks and show that it can successfully repair all of them. Also, compared to related techniques, AuRUS can produce many unique solutions while showing more scalability.

Background: Test flakiness is identified as a major issue that compromises the regression testing process of complex software systems. Flaky tests manifest non-deterministic behaviour, send confusing signals to developers, and break their trust in test suites. Both industrial reports and research studies highlighted the negative impact of flakiness on software quality and developers' productivity. While researchers strive to devise solutions that could help developers addressing test flakiness, the elaboration and assessment of these solutions are hindered by the lack of datasets large enough to leverage learning techniques. Aim: To address this lack, we conduct an exploratory study that investigates a new mean for producing datasets of flaky tests. Method: We propose an approach that relies on program mutation to inject flakiness in software tests. In particular, we plan to delete helper statements from tests to make their outcomes order-dependent, i.e., pass in certain running orders but fail in other orders. We intend to apply our mutation-based approach to a set of 14 Java projects to assess the effectiveness of test mutation in injecting order-dependency and generate a new dataset that could be used to study flakiness.

Vulnerability prediction refers to the problem of identifying system components that are most likely to be vulnerable. Typically, this problem is tackled by training binary classifiers on historical data. Unfortunately, recent research has shown that such approaches underperform due to the following two reasons: a) the imbalanced nature of the problem, and b) the inherently noisy historical data, i.e., most vulnerabilities are discovered much later than they are introduced. This misleads classifiers as they learn to recognize actual vulnerable components as non-vulnerable. To tackle these issues, we propose TROVON, a technique that learns from known vulnerable components rather than from vulnerable and non-vulnerable components, as typically performed. We perform this by contrasting the known vulnerable, and their respective fixed components. This way, TROVON manages to learn from the things we know, i.e., vulnerabilities, hence reducing the effects of noisy and unbalanced data. We evaluate TROVON by comparing it with existing techniques on three security-critical open source systems, i.e., Linux Kernel, OpenSSL, and Wireshark, with historical vulnerabilities that have been reported in the National Vulnerability Database (NVD). Our evaluation demonstrates that the prediction capability of TROVON significantly outperforms existing vulnerability prediction techniques such as Software Metrics, Imports, Function Calls, Text Mining, Devign, LSTM, and LSTM-RF with an improvement of 40.84% in Matthews Correlation Coefficient (MCC) score under Clean Training Data Settings, and an improvement of 35.52% under Realistic Training Data Settings., Comment: The article was accepted in Empirical Software Engineering (EMSE) on July 02, 2022

Graph-based Semi-Supervised Learning (GSSL) is a practical solution to learn from a limited amount of labelled data together with a vast amount of unlabelled data. However, due to their reliance on the known labels to infer the unknown labels, these algorithms are sensitive to data quality. It is therefore essential to study the potential threats related to the labelled data, more specifically, label poisoning. In this paper, we propose a novel data poisoning method which efficiently approximates the result of label inference to identify the inputs which, if poisoned, would produce the highest number of incorrectly inferred labels. We extensively evaluate our approach on three classification problems under 24 different experimental settings each. Compared to the state of the art, our influence-driven attack produces an average increase of error rate 50\% higher, while being faster by multiple orders of magnitude. Moreover, our method can inform engineers of inputs that deserve investigation (relabelling them) before training the learning model. We show that relabelling one-third of the poisoned inputs (selected based on their influence) reduces the poisoning effect by 50\%.

Much research on software engineering and software testing relies on experimental studies based on fault injection. Fault injection, however, is not often relevant to emulate real-world software faults since it "blindly" injects large numbers of faults. It remains indeed challenging to inject few but realistic faults that target a particular functionality in a program. In this work, we introduce IBIR, a fault injection tool that addresses this challenge by exploring change patterns associated to user-reported faults. To inject realistic faults, we create mutants by retargeting a bug report driven automated program repair system, i.e., reversing its code transformation templates. IBIR is further appealing in practice since it requires deep knowledge of neither of the code nor the tests, but just of the program's relevant bug reports. Thus, our approach focuses the fault injection on the feature targeted by the bug report. We assess IBIR by considering the Defects4J dataset. Experimental results show that our approach outperforms the fault injection performed by traditional mutation testing in terms of semantic similarity with the original bug, when applied at either system or class levels of granularity, and provides better, statistically significant, estimations of test effectiveness (fault detection). Additionally, when injecting 100 faults, IBIR injects faults that couple with the real ones in 36% of the cases, while mutants from mutation testing inject less than 1%. Overall, IBIR targets real functionality and injects realistic and diverse faults.

An established way to improve the transferability of black-box evasion attacks is to craft the adversarial examples on an ensemble-based surrogate to increase diversity. We argue that transferability is fundamentally related to uncertainty. Based on a state-of-the-art Bayesian Deep Learning technique, we propose a new method to efficiently build a surrogate by sampling approximately from the posterior distribution of neural network weights, which represents the belief about the value of each parameter. Our extensive experiments on ImageNet, CIFAR-10 and MNIST show that our approach improves the success rates of four state-of-the-art attacks significantly (up to 83.2 percentage points), in both intra-architecture and inter-architecture transferability. On ImageNet, our approach can reach 94% of success rate while reducing training computations from 11.6 to 2.4 exaflops, compared to an ensemble of independently trained DNNs. Our vanilla surrogate achieves 87.5% of the time higher transferability than three test-time techniques designed for this purpose. Our work demonstrates that the way to train a surrogate has been overlooked, although it is an important element of transfer-based attacks. We are, therefore, the first to review the effectiveness of several training methods in increasing transferability. We provide new directions to better understand the transferability phenomenon and offer a simple but strong baseline for future work., Comment: Accepted at UAI 2022

The rapid spread of the Coronavirus SARS-2 is a major challenge that led almost all governments worldwide to take drastic measures to respond to the tragedy. Chief among those measures is the massive lockdown of entire countries and cities, which beyond its global economic impact has created some deep social and psychological tensions within populations. While the adopted mitigation measures (including the lockdown) have generally proven useful, policymakers are now facing a critical question: how and when to lift the mitigation measures? A carefully-planned exit strategy is indeed necessary to recover from the pandemic without risking a new outbreak. Classically, exit strategies rely on mathematical modeling to predict the effect of public health interventions. Such models are unfortunately known to be sensitive to some key parameters, which are usually set based on rules-of-thumb.In this paper, we propose to augment epidemiological forecasting with actual data-driven models that will learn to fine-tune predictions for different contexts (e.g., per country). We have therefore built a pandemic simulation and forecasting toolkit that combines a deep learning estimation of the epidemiological parameters of the disease in order to predict the cases and deaths, and a genetic algorithm component searching for optimal trade-offs/policies between constraints and objectives set by decision-makers. Replaying pandemic evolution in various countries, we experimentally show that our approach yields predictions with much lower error rates than pure epidemiological models in 75% of the cases and achieves a 95% R2 score when the learning is transferred and tested on unseen countries. When used for forecasting, this approach provides actionable insights into the impact of individual measures and strategies.

We introduce SeMu, a Dynamic Symbolic Execution technique that generates test inputs capable of killing stubborn mutants (killable mutants that remain undetected after a reasonable amount of testing). SeMu aims at mutant propagation (triggering erroneous states to the program output) by incrementally searching for divergent program behaviours between the original and the mutant versions. We model the mutant killing problem as a symbolic execution search within a specific area in the programs' symbolic tree. In this framework, the search area is defined and controlled by parameters that allow scalable and cost-effective mutant killing. We integrate SeMu in KLEE and experimented with Coreutils (a benchmark frequently used in symbolic execution studies). Our results show that our modelling plays an important role in mutant killing. Perhaps more importantly, our results also show that, within a two-hour time limit, SeMu kills 37% of the stubborn mutants, where KLEE kills none and where the mutant infection strategy (strategy suggested by previous research) kills 17%.